The Business of Making People Compete to Work

I started looking at Vibe Jam because I wanted to understand how it made money. On the surface, it is a game competition. People use AI tools to build strange little games, sponsors supply prizes, winners receive attention, and the internet gets a few weeks of playable chaos. The obvious business model is sponsorship. Cursor and other AI companies pay to be placed in front of developers who might use their products.

But the more interesting part is what the sponsors receive in return. They are not simply buying banners on a website. They are buying a temporary wave of people installing their tools, reading their documentation, testing their limits, producing demonstrations and publishing the results. The event turns product adoption into a competition. Instead of paying developers one by one to experiment with an API, the sponsor helps create a situation in which hundreds of developers compete for the opportunity.

ETHGlobal runs a more institutional version of the same machine. Blockchain protocols fund prize tracks that reward teams for building with particular networks, wallets, identity systems and infrastructure. The prizes are presented as open possibilities, but they also act as directional signals. A sponsor can place $10,000 or $20,000 at the end of a technical path, and developers will spend a weekend learning everything required to reach it.

It is tempting to describe this as a clever new form of developer marketing. That is true, but it misses the longer history. Vibe Jam and ETHGlobal belong to a wider competition business that has been developing for decades. Kaggle does it with machine learning. Topcoder does it with software and design. InnoCentive and HeroX do it with scientific and industrial problems. XPRIZE does it with large technical missions. Governments use challenge prizes for spaceflight, health, energy and public services. Companies use internal competitions to search for ideas that their own organizational structures have failed to produce.

The recurring idea is simple. When an organization does not know who can solve a problem, how long the solution should take or what the best approach will look like, it can avoid choosing a team in advance. It publishes the problem, offers a reward and lets many people absorb the uncertainty.

Kaggle may be the clearest example. A company or research organization provides a dataset and a measurable objective. Thousands of participants may then test different models, features and methods. The leaderboard continuously ranks their performance. Only a small number will receive prize money, but the organizer benefits from the search conducted by everyone.

A normal employer chooses a person, pays for their time and hopes that their approach works. A competition pays for the result after comparing it with many alternatives. Most of the cost of failure remains with the participants.

That changes the economics of experimentation. If a company hired 500 machine-learning researchers to independently attack the same problem, it would be absurdly expensive. If it puts the problem on Kaggle, offers a prize and attracts 500 participants, the total amount of work may be similar while the company’s direct financial commitment is dramatically smaller. Participants accept this because they are also competing for reputation, practice, career opportunities and proof that they belong among the best.

The leaderboard is important because it makes invisible work visible. A participant does not need a prestigious university, a polished résumé or the correct professional network. Their model either performs well or it does not. This can create real mobility. An unknown person can defeat experienced researchers. A small team can outperform a corporate laboratory. Competitions sometimes reveal talent that conventional hiring systems would miss.

They also create a large amount of unpaid or underpaid work around the winning result. Every failed model contributes to the search, even when the organization never sees the details. Every participant spends time learning the problem, cleaning the data, understanding the domain and testing approaches. The prize does not need to compensate the total labor because the participants are purchasing other things with their time: education, status, community, hope and the possibility of a different career.

Topcoder applies a related model to software production. Companies divide work into challenges and invite a distributed crowd to compete on algorithms, interfaces, development tasks or technical solutions. The customer gains access to a workforce without building a conventional team around every problem. The platform handles the rules, submissions, evaluation and payments.

InnoCentive extended this logic into corporate research and science. Organizations could publish problems that had resisted their internal teams and offer rewards to outside solvers. The person with the answer might be working in a different industry, living in another country or applying a technique the original company would never have considered. The competition becomes a search mechanism for knowledge that exists somewhere but cannot be found through the organization’s normal boundaries.

HeroX and XPRIZE make the same mechanism more public and theatrical. A large challenge can attract teams, investors, universities and media attention around a problem that previously belonged to a small group of specialists. The prize gives the problem a narrative. There is a starting point, a finish line and a visible reward. Technical work becomes a race that outsiders can understand.

The prize amount can also be misleading. A $10 million prize sounds enormous until the participating teams collectively spend several times that amount trying to win it. The organizer does not necessarily finance the full cost of solving the problem. It creates a credible reward that persuades other people to finance their own attempts.

This may be one of the most powerful properties of competitions. They mobilize resources that the organizer does not own. Participants contribute their own time. Teams find their own investors. Universities provide laboratories. Companies lend equipment. Volunteers create communities around the challenge. The prize becomes a coordination device for a much larger pool of effort.

Vibe Jam is a small and culturally current version of this older structure. The problems are playful, the projects can be produced quickly, and the participants already have access to inexpensive AI tools. There is no need to build a laboratory or spend years developing a spacecraft. A person can join from a laptop, make a game in a few days and post it online.

This makes the model especially efficient as marketing. The participants create the product demonstrations, launch posts, videos, memes and social distribution. Each project becomes evidence that the sponsor’s tool can produce something. Even a broken game may still show an interesting technique. Even a losing submission can teach other people how the underlying tools work.

The audience is performing many of the functions that would previously have been divided among an advertising agency, a developer-relations team, a research department and a recruitment firm. They test the product, explain it, give it cultural meaning and identify the strongest builders. The sponsor pays for the environment in which this happens.

I do not think the participants are necessarily being tricked. Many of them know exactly what the arrangement is. They may genuinely enjoy the event. They may learn more in one weekend than they would from weeks of tutorials. They may meet collaborators, produce a portfolio project or discover that they are good at something. A competition can create a level of concentration that ordinary work rarely achieves.

Still, the distribution of value is unusual. The organizer and sponsor benefit from the total field of activity, while rewards are concentrated among a few winners. The participants collectively create far more value than the prize pool distributes. This imbalance is tolerated because each person evaluates the event individually. They are not asking whether the entire crowd is being fairly compensated. They are asking whether spending one weekend gives them a worthwhile chance at learning, recognition or escape.

That is why competition businesses can feel generous and extractive at the same time. They open doors that conventional institutions keep closed, but they also convert ambition into subsidized production. They allow outsiders to participate, while benefiting from the fact that there will always be more hopeful outsiders than available prizes.

The platform sits in the most durable position. Kaggle does not need to employ every data scientist. Topcoder does not need to maintain a permanent team for every client problem. ETHGlobal does not need to build every blockchain application. Vibe Jam does not need to develop the games. The platform owns the structure in which other people’s work becomes searchable, comparable and commercially useful.

Over time, it also accumulates data about the participants. It learns who enters, who finishes, who performs well, which tools they use, which teams form repeatedly and which projects continue afterward. That information can support recruiting, investment, education, grants and future competitions. The visible event may last a weekend, but the database of ambition keeps growing.

The most defensible asset may therefore be the ability to repeatedly create credible contests. A successful competition requires more than publishing a prize. Participants need to believe the organizer can attract attention, judge fairly, distribute rewards and make winning matter. Sponsors need to believe the crowd is talented and relevant. Each successful event strengthens this trust, making the next one easier to sell.

This is why ETHGlobal is deeper than a series of hackathons and why Kaggle became more than a website with leaderboards. They have built reputational infrastructure around competition. A company arrives with a problem or a product. The platform turns it into a challenge that a community considers worth solving.

I can imagine this business model expanding as AI makes production cheaper. When software, visual design, video, research and analysis can all be produced faster, it becomes possible to run competitions around increasingly specific commercial problems. A company might ask thousands of people to redesign one customer journey, find new uses for a dataset, develop sales campaigns, analyze a scientific literature or create interfaces for an unreleased device.

The outputs would not need to be immediately usable. The purpose might simply be to map the possibility space. An organization could observe hundreds of approaches before deciding which direction deserves serious investment. The competition would function as distributed pre-research.

This could be valuable for public problems too. Cities could run competitions around transportation, healthcare access or bureaucratic services. Laboratories could publish carefully structured diagnostic problems. Museums could invite people to reinterpret their collections. Schools could organize students around local challenges.

But the quality of the outcome would depend on who defines the metric. Kaggle competitions optimize what the leaderboard can measure. Hackathons optimize what judges can understand during a short presentation. XPRIZE teams optimize against the conditions required to win. Once a problem becomes a competition, the scoring system begins to shape the solution.

That may be the central tension. Competitions can unlock enormous amounts of distributed intelligence, but they also narrow attention toward whatever can be judged, ranked and rewarded. Important work that is slow, ambiguous or difficult to demonstrate may disappear from view. A good competition produces a clear winner. A real problem may resist having one.

Vibe Jam looks like people making AI games for fun, but it sits inside a much larger transformation in how organizations acquire work. Instead of assembling a team and directing it through employment, a company can publish an opportunity and attract a crowd. It pays a few winners, observes everyone else and retains the broader benefits of the activity.

The company of the future may not need to employ all the people working on its problems. It may only need to own the prize, the leaderboard and the place where ambitious people gather.

Connected reading

• The Product Is a Decoy
• The Product Happens Between the Screens
• The Problems Companies Admit They Have
• Why Is AI Making Me Think About Business Strategy?

AI authorship note

This essay began with Kris Haamer asking how Cursor Vibe Jam makes money, then noticing that its sponsor-funded structure resembles ETHGlobal. Kris pushed the argument outward to include Kaggle, Topcoder, InnoCentive, HeroX and XPRIZE, turning the article into an exploration of the wider competition economy rather than a narrow piece about hackathons. AI helped research the organizations, connect the business models, develop the labor and incentive argument, edit the essay and prepare the Raw Drafts publishing files. Kris supplied the original questions, the wider competition framing, the editorial direction and the final decision to publish.